ISSUE 329   ·   March 30, 2021

In the News

Matrix Multiplication Inches Closer to Mythic Goal

A recent paper set the fastest record yet for multiplying two matrices. But it also marks the end of the line for a method researchers have relied on for decades to make improvements.
Quanta Magazine | Kevin Hartnett

Platforms vs. PhDs: How tech giants court and crush the people who study them

Although scraping public websites may be legal, Big Tech writes the rules for its own platforms. Some of those rules are being challenged in courts but meanwhile, the dynamic between Big Tech platforms and the people who study how those platforms use data is rife with tension.
Protocol | Issie Lapowsky

Sponsored Link

How to get to a higher-performing model faster, with active learning

Curious what active learning could do for your model? Alegion breaks down the three major advantages to using active learning as part of your model development. Alegion helps customers with their model development when models need to get better and better at granular classification of low confidence instances to get to higher quality, faster. Learn more.

Tutorials, Projects & Opinions

How We Built a Context-Specific Bidding System for Etsy Ads

This is a great post that walks through the considerations and algorithms of Etsy's online ad platform. Here's how they moved from making batch predictions once per day to dynamic predictions up to 12,000 times per second. The post is easy to follow and gets into details that aren't often shared.
Etsy | Code as Craft

Unit testing Python code in Jupyter notebooks

Whether you are a unit testing purist or you just want to sprinkle a few tests into your notebooks, here are a few options to consider. Includes links and sample code for each approach.
Matt Wright

The ghosts in the data

Implicit knowledge is knowledge that exists within expert communities but isn't written down. To be effective, it's key to understand what that knowledge is and how to access it. In her latest post, Vicki Boykis explores this "ghost" knowledge in the data world.
Vicki Boykis

GPT-3 Powers the Next Generation of Apps

GPT-3 launched just nine months ago and already, it powers more than 300 applications with semantic search, summarization, sentiment analysis, content generation, translation, and more. Here's an update from the project including platform improvements and a selection of applications that show its range of capabilities.
OpenAI

R vs. Python vs. Julia

This article isn't about which is better. If you want to write code that's efficient, which language should you choose?
Daniel Moura

Code & Tools

Create Your Own VS Code Themes

Create your own or install a theme from the gallery with this theme builder for VS Code.
Theme Studio

Data Visualization

Dimension reduction 1

This set of slides is an awesome introduction to Principal Components Analysis (PCA). The examples are easy to follow, there are worthwhile references at the end and, for R users, there's sample code too.
Claus O. Wilke

Elevation Scan

Elegant approach for visualizing max/mean elevation across the world. This is a work of art in code.
Observable | Nicolas Lambert

Sign up to get Data Elixir's  data science newsletter in your Inbox >>

Data Elixir is curated and maintained by Lon Riesberg. If you have questions or suggestions for the newsletter, just reply back to this email.